Spatial structure of the plasma sheet boundary layer at distances greater than 180 RE as derived from energetic particle measurements on GEOTAIL

We have analyzed the onsets of energetic particle bursts detected by the ICS and STICS sensors of the EPIC instrument on board the GEOTAIL spacecraft in the deep magnetotail (i.e., at distances greater than 180 RK). Such bursts are commonly observed at the plasma-sheet boundary layer (PSBL) and are highly collimated along the magnetic field. The bursts display a normal velocity dispersion (i.e., the higher-speed particles are seen first, while the progressively lower speed particles are seen later) when observed upon entry of the spacecraft from the magnetotail lobes into the plasma sheet. Upon exit from the plasma sheet a reverse velocity dispersion is observed (i.e., lower-speed particles disappear first and higher-speed particles disappear last). Three major findings are as follows. First, the tailward-jetting energetic particle populations of the distant-tail plasma sheet display an energy layering: the energetic electrons stream along open PSBL field lines with peak fluxes at the lobes. Energetic protons occupy the next layer, and as the spacecraft moves towards the neutral sheet progressively decreasing energies are encountered systematically. These plasma-sheet layers display spatial symmetry, with the plane of symmetry the neutral sheet. Second, if we consider the same energy level of energetic particles, then the H ‘ layer is confined within that of the energetic electron, the He++ layer is confined within that of the proton, and the oxygen layer is confined within the alpha particle layer. Third, whenever the energetic electrons show higher fluxes inside the plasma sheet as compared to those at the boundary layer, their angular distribution is isotropic irrespective of the Earthward or tailward character of fluxes, suggesting a closed field line topology.


Introduction
Although single-point measurements cannot distinguish spatial from temporal processes, in this paper we present evidence for spatial layering of the deep-tail plasma sheet at distances greater than 180 R E .
The plasma-sheet boundary layer (PSBL) is the region between the plasma sheet and the tail lobes, comprised of highly anisotropic ion distributions (Lui et al., 1978;DeCoster and Frank, 1979;Mobius et al., 1980;Spjeldvik and Fritz, 1981;Williams, 1981;Eastman et al., 1984Eastman et al., , 1985Sarris, 1990, 1991a, b). In the distant tail the PSBL represents the layer of reconnected ®eld lines lying between the separatrix and the thermal plasma sheet (Richardson and Cowley, 1985). Tailward-streaming cold ion beams of ionospheric origin at energies less than 1 keV are also frequently observed in the PSBL (Parks et al., 1984;Eastman et al., 1984).
The PSBL¯uxes are expected and observed to E´B drift toward the neutral sheet for a given externally imposed cross-tail electric ®eld E (Sarris and Axford, 1979;Williams, 1981;Andrews et al., 1981;Richardson and Cowley, 1985;Scholer et al., 1987). The cross-tail ®eld causes two streams of dierent energy ions ejected at zero pitch angles from an acceleration source to drift toward the neutral sheet with dierent drift V drift , so that the beams become separated in space. Mobius et al. (1980) performed a case-study of the near-Earth PSBL during substorm recovery. They found that the particle spectra at the central plasma sheet (CPS) were softer than at the PSBL, and that the alpha particle layer was con®ned to within the proton layer. Their result will be extended in this work for the distanttail plasma sheet.
Close to Earth, although highly anisotropic angular distributions are often encountered at the PSBL, such energetic particle populations are typically observed at very low intensities relative to the CPS Sarris, 1990, 1991a, b). Only in one exceptional case (Sarafopoulos and Sarris, 1991a) at $37 R E downtail, did the higher¯uxes of energetic protons occur at the PSBL. Conversely, in the distant plasma sheet (as we shall see in this work), it is not uncommon that peak uxes of streaming particles occur at the PSBL. In such cases, the¯uxes are associated with open magnetic ®eld topology. Richardson and Cowley (1985) examined tailwardstreaming ion bursts extending into the lobe for a few minutes before entry into or after exit from the plasma sheet in the deep geomagnetic tail. They showed evidence at the leading edge of a normal velocity dispersion (i.e., high-energy particles are detected before low-energy ones), whereas at the trailing edge a reverse dispersion was detected. Such features have been recognized as the result of spatially separated¯uxes. Similarly, Scholer et al. (1986Scholer et al. ( , 1987 and Scholer (1986) found that the velocity dispersion eect during the appearance and disappearance of the beams in the deep PSBL indicates that these beams are not due to temporal eects. Our work concurs with these authors' conclusions and goes further to demonstrate the substructure of the distant-tail PSBL using GEOTAIL data sets.

Observations and data analysis
We selected three intervals for thorough study. The ®rst interval encompasses seven successive PSBL crossings and helps demonstrate the symmetry of the energetic particle layers with respect to the neutral sheet. The intervals are 16:00±19:30 UT on 14 March (day 73) 1993, and 08:28±08:43 and 14:25±14:45 UT on 18 April (day 108) 1994. We used data from the magnetic-®eld (MGF) experiment (Kokubun et al., 1994) at 3-s resolution and the energetic particle (EPIC) instrument (Williams et al., 1994). EPIC consists of sensors ICS and STICS, and the time resolution is declared in each plot varying from 3 to 96 s. All the EPIC measurements are depicted as dierential¯uxes.

Overview of studied interval: 14 March (day73) 1993
An overview plot of the period 16:00±19:30 UT on 14 March 1993 is presented in Fig. 1. At 18:00 UT the GEOTAIL spacecraft was located at (X, Y, Z ) = (A177, 35.4, 4.6) R E in GSE coordinates. The third panel shows the B x and B z components of the vector magnetic ®eld in GSM coordinates. It is obvious that during this interval the entire magnetotail apparently moved northward and then southward, causing GEOTAIL to pass from the northern lobe into the plasma sheet (16:52 UT), then into the southern-hemisphere lobe (17:24), and ®nally back into the northern lobe via the plasma sheet (18:10±18:25 UT). A brief excursion of the spacecraft into the CPS (17:49±17:55 UT) and a ®nal entry into the plasma sheet (18:52 UT) are also notable. The H + and He ++ dierential intensities of the EPIC/STICS data sets (second and ®fth panels) reveal prevailing tailward¯uxes (solid circles), as well as the existence of an interval of intense Earthward¯ux (open triangles) simultaneously with the second passage throughout the plasma sheet. Energetic (>32 keV) electrons of the EPIC/ICS data are shown in the ®rst panel, demonstrating that higher-streaming¯uxes occurred preferentially at the plasma sheet adjacent to the lobe magnetic ®eld or on lobe ®eld lines. The same is valid with all species of energetic uxes, and this will become clearer in the next paragraphs through a systematic study of all the spacecraft transitions. A representative¯ux of 58±77-keV energetic protons is displayed in the fourth panel, and a discernible time-lag of $4 min is apparent between electrons and protons for the ®rst detected onsets of uxes. The distinct spatial separation among high-and lowenergy protons (shown at the bottom panel of Fig. 2) is further illustrated with the help of energy spectra shown in Fig. 3. The PSBL and CPS are identi®ed as the layers where the magnetic ®eld magnitude is greater than 7 nT and less than 4 nT, respectively. The proton energy spectrum at the CPS on the basis of EPIC/STICS measurements is softer than that at the PSBL. An explanation for the softening of the spectrum is given in terms of a cross-tail electric ®eld causing dierent energy beams of protons to drift inward with the same speed, so that the beams become separated in space.
The spatial structure of the plasma sheet is studied in detail in Fig. 4a, which corresponds to the GEOTAIL transition from the north lobe to the CPS. We observe in good order time-delayed peak¯uxes of energetic protons (tailward¯uxes with 24-or 48-s resolution) relative to energetic electrons (tailward¯uxes with 3-s resolution). The energies of protons continuously decrease toward the bottom panel from superthermal to thermal population (only 9.4 keV). It is evident that the plasma sheet is itself composed of successive layers of ejected particles of decreasing energy as the spacecraft moves from the separatrix and approaches the CPS. Figure 4b shows in detail the reverse velocity dispersion detected as the spacecraft moved from almost the neutral sheet to the south lobe. The peak¯ux at 9.4-keV protons occurred at the innermost depths of the plasma sheet (see also Fig. 2). This reverse dispersion can be interpreted only in terms of spatial separation of the ejected particles from the acceleration source. In addition, the preceding observations imply that the plasma sheet displays a symmetric structure with the neutral sheet being the plane of symmetry. The second and bottom panels give the He + + and proton¯uxes, respectively, at the same time for the lowest measured energy level ($9.4 keV, solid lines ) and the much higher one (greater than 55 keV, dashed lines ) Fig. 3. The central plasma sheet shows a softer spectrum for energetic protons as compared to that at the plasma sheet boundary layer. This ®gure is derived from data corresponding to the interval 17:05±17:25 UT of Fig. 2 Fig. 4.a. We zoom in on the ®rst GEOTAIL entry interval into the plasma sheet (see Fig. 2). All the panels show the tailward (dominant) dierential¯uxes. The top panel gives the >32-keV electrons with resolution 3 s from EPIC/ICS. The second panel gives the 155±228-keV¯ux of protons with resolution 48 s from EPIC/ICS. The next panels show the 55.7, 22.9, 14.6, and 9.4-keV¯uxes of protons with resolution 24 s obtained via the EPIC/STICS sensor. The systematic displacement of peak¯uxes (i.e., the normal velocity dispersion eect) is apparent. b. Format is the same as a. The¯uxes show systematic reverse velocity dispersion as we zoomed in on the ®rst exit interval from the plasma sheet (see Fig. 2). c From top to bottom spinaveraged dierential¯uxes of >32-keV electrons (resolution 3 s), 187±222-keV of the CNO group (resolution 6 s), 58±77-kev protons (resolution 6 s), 194±281-keV alphas (resolution 48 s), 77±107-keV protons (resolution 96 s), 70±96-keV alphas (resolution 6 s). The dierent delay times of peak¯uxes are due to a velocity ®lter eect and re¯ect the spatial structure of the plasma sheet Figure 4c further exempli®es the eect of velocity dispersion using dierent species. It is evident that: 1. When the alpha particles have almost four times the energy of protons then the same arrival time is anticipated for both. This is the situation displayed in third and fourth panels. 2. The bottom two panels show that the alpha particles lag in arrival relative to protons, and this is anticipated because both of them have nearly equal total energies. Therefore, the alpha layer is con®ned to within that of the protons. 3. The delay time of the $187-keV oxygen particles relative to electrons is 10 min and that of the $70-keV alphas is 8 min (i.e., ratio 1.25) in expected relative delay based on the speeds of these species (i.e., ratio 1.22). 4. The $58-keV protons and the $70-keV alpha particles (third and sixth panels in Fig. 4c) give 4-and 8min observational delay times, respectively, relative to electrons. The ratio of these delays is very close to the theoretically expected ratios for the species (that is, 0.548).
Thus, on the basis of a, b and c Fig. 4 the plasma sheet could be regarded as``highly strati®ed with a multitude of successive layers''. Certainly the word`l ayer'' here corresponds to the number of our channels. As a matter of fact, the particles are spatially ordered according to a velocity ®lter eect. The velocity ®lter eect results in a continuous splitting of particles originating from a source producing particles with dierent velocities. Nevertheless, a sketch of ®ve successive energy-dependent layers are schematically shown in Fig. 5 in association with the strength of the local magnetic ®eld. The distance scale in ®gure is arbitrary, because we fail to determine the velocity of the spacecraft relative to the PSBL (that would be derived from gradient anisotropy of energetic particles). Indicative angular distributions of the energetic particles at the PSBL as well as at the CPS are displayed and discussed later on.
2.3 A brief excursion from south lobe toward the neutral sheet (17:49±14:55 UT) Figure 6 shows a gradual GEOTAIL excursion from the south-lobe magnetic ®eld toward the CPS (see third panel). In general, we observe that the higher¯uxes of energetic electrons (®rst panel), as previously, occurred at higher magnetic ®eld values, whereas the more highly energetic ion intensities detected essentially with relatively lower magnetic ®eld magnitudes and therefore closer to the neutral sheet. The electrons display tailward beam-like angular distributions adjacent to the lobe plasma sheet, while at the CPS regions they are isotropic. The dierentiation between tailward (Tw) and Earthward (Ew)¯uxes is denoted with solid-dotted and dashed lines, respectively. The 35±46.8-keV ions show a $30-s lag in onset, while they lead the way by 60 s in exiting the plasma sheet. This dierence could be associated with the more gradual entry into the plasma

Additional examples of bursts with dispersive onsets
Two more examples of dispersive onsets of energetic particle bursts, which occurred on 18 April (day 108) 1994 are shown in Figs. 9 and 10. The two ®gures correspond to an entry into and an exit from the plasma sheet, respectively. In the former case a normal velocity dispersion in peak¯uxes is apparent and spans a period of $10 min, whereas in the latter case a reverse velocity dispersion covers a period of more than 4 min. On both cases all the species stream tailward. The spacecraft was located at (X, Y, Z ) GSM = (A197.6, 21.5, 1.8) and (X, Y, Z ) GSM = (A197.5, 22.1, 1.5) during the intervals of Figs. 9 and 10, respectively.

Discussion
The main purpose in this article is to establish the highly strati®ed spatial structure of the distant plasma sheet with few representative examples. Our results con®rm previous ISEE-3-examined velocity dispersions also interpreted as spatial features. The temporal resolution, species dierentiation, and multiplicity of channels of the EPIC instrument provide us with a further insight into the problem of structure and dynamics of the Earth's magnetotail. Indeed, in the foremost case-study of this work, the spatial structure of the plasma sheet is monitored in successive snapshots, as an active X-line was possible propagating over the spacecraft. The successive entries into and exits from the plasma sheet (seven crossings) associated with normal or reverse velocity dispersions, give strong evidence of a quasipermanent plasma-sheet structure for at least 2.5 h. Had the detected velocity dispersions been considered a result of varying times of¯ight of the dierent velocity particles ejected from an X-line, the following problems would have arised: 1. The reverse velocity dispersion during exits from the plasma sheet would not be explainable by the same process. 2. A smooth and gradual entry into the plasma sheet is associated with sustained¯uxes of energetic electrons (see $17:10 UT in Fig. 6). Instead, an abrupt exit is associated with short-lived¯uxes. In the latter case the electron layer is traversed brie¯y (see $17:54 UT in Fig. 6).  3. Local peaks of energetic electrons are observed to be well associated with local magnetic ®eld increases showing the spacecraft approach to the PSBL. 4. Successive peaks of¯uxes of tailward-streaming electrons occur systematically and exclusively adjacent to the lobe plasma sheet or on lobe ®eld lines, while the spacecraft traverses the whole plasma sheet. 5. The low-energy particles (i.e., 9.4 keV) occur in essence at almost zero magnetic ®eld magnitudes. Richardson and Cowley (1985) examined strongly tailward-streaming ion bursts extending into the lobe for a few minutes after exit from the plasma sheet at the deep geomagnetic tail. They found evidence at the leading edge a normal velocity dispersion (i.e., highenergy particles are detected before low-energy ones), whereas at the trailing edge a reverse dispersion was detected, and such features have been recognized as the result of spatially separated¯uxes. Similarly, Scholer et al. (1987) found that the velocity dispersion eect during the appearance and disappearance of the beams in the deep PSBL indicates that these beams are not due to temporal eects. Indeed, this work agrees with such a conclusion and goes further to demonstrate with an improved data set the substructure of the PSBL in the distant plasma sheet. Parks et al. (1992) examined the PSBL crossings close to the Earth ($20 R E downtail) at 0.16-s resolution, and demonstrated that the proton and electron boundaries of energies 2±6 keV are displaced in time. This feature was interpreted as indicative of a genuine spatial structure not caused by the dierent Larmor radii of electrons and protons, since then the ions would be extended further into the lobe. Instead, what was observed was that the electron boundary is extended further out into the lobe.
At the PSBL close to the Earth, it is also well documented that although highly anisotropic angular distributions are often encountered, nevertheless in the vast majority of the cases they are associated with extremely low energetic particle intensities (Sarafopoulos and Sarris, 1990Sarris, , 1991a. Only one exceptional case-study has been presented where at $37 R E downtail simultaneous beam-like angular distributions at the PSBL and higher¯uxes of energetic ions compared to those at the CPS were observed (Sarafopoulos and Sarris, 1991a). The situation is dierent for the PSBL at distances greater than 180 R E , as we have concluded in the present article. The deep-tail PSBL frequently displays peak¯uxes of energetic ions and electrons, while weaker¯uxes occur as the spacecraft moves inward to innermost depths of the plasma sheet. In this work the majority of PSBL crossings associated with tailward-streaming population comply with such a result. However, PSBL crossings exhibiting the opposite behavior are also detected. The distinct feature of the latter crossings is that the higher¯uxes are associated with isotropic electrons. And in such a closed topology of the magnetic ®eld the trapped particles have all the required time to convect inward, creating peak¯uxes at the CPS (see Fig. 7). To support observationally the last statement we present an additional striking case asso-ciated with Fig. 11. In this example the more intensē uxes of energetic 58±77-keV protons occurred at the CPS but with isotropic electrons. A close examination of Fig. 11 reveals that although the energetic protons display a tailward¯ux, the energetic electrons are isotropic. Conversely, at 10:20 UT the high¯uxes of energetic electrons that occurred at the PSBL are associated with highly anisotropic angular distributions.
In the near-Earth magnetotail the direction of the bulk plasma¯ow, as well as the energetic particle¯uxes, varies considerably. Conversely, in the deep magnetotail (distances greater than 150 R E ) the tailward population becomes the dominant characteristic, while Earthward ows occurred in less than 1% of all cases (Zwickl et al., 1984;Daly et al., 1984;Schindler et al., 1989). In the present work the reversal to Earthward¯uxes of all ions observed at $180 R E downtail (for an interval of $12 min on 14 March 1993) could be well interpreted in terms of a neutral line retreating tailward over the spacecraft. Figure 12 presents common-scale (200 nT per division) ground magnetograms from auroral stations. On x-component traces substorms can be identi-®ed by sharp negative perturbations (westward electrojets) or by positive excursions (eastward electrojets). Critical in the identi®cation of substorm onset is the local time of the station. Only stations that are close to midnight (2 h of local time) can be used for substorm identi®cation. Digital ground magnetometer coverage is not very good in the Russian sector and therefore an accurate determination of substorm onset based on the limited number of nightside stations between 15:00 and 23:00 UT is not always possible. At $16:48 UT a negative bay onset at the Russian station TIX started, and this time could be considered as the substorm onset given that for the Tixie station the MLT = 0 occurs at 15:37 UT. At the same time the Fig. 11. The >32-keV energetic electrons show two peak¯uxes (second panel ). The ®rst occurred during the entry into the plasma sheet (third panel ) and displays tailward¯ux with beam-like angular distributions, while the second occurred at the CPS and shows isotropic electron¯uxes. The 58±77-keV protons (®rst panel ) convect tailward (thick line ) positive excursion at the Leirvogur observatory (LRV) is produced by the eastward electrojet, because for this station the MLT = 0 occurs at 00:15 UT. The Earth-ward¯ux detected at GEOTAIL during the interval 18:10±18:25 UT could well be associated with the recovery phase of this concrete substorm that started at $16:48 UT. The earliest detected energetic electrons were measured at $16:52 UT, that is, only $4 min after the substorm onset. The near-Earth neutral-line model for substorms suggests that in this case a new acceleration site must be formed in the near-Earth region (e.g., at A20R E ) at $16:48 UT, and predicts that a plasmoid must be formed in the near-Earth tail and subsequently propagated downtail. Indeed, a possible interpretation of the phenomenon of tailward-to-Earthward¯ow reversal may be the tailward retreat of a plasmoid downstream of spacecraft. The time-delay between the substorm onset and the appearance of Earthward¯ux is $82 min. If the downtail propagation velocity of the acceleration site is constant, then the speed is $205 km s A1 (i.e., 160 R E divided by 82 min). Angelopoulos et al. (1996) found a speed 75±115 km s A1 out of 14 substorms, while GEOTAIL was located at A90 R E .
If we compare the two results then the assumption of constant speed of plasmoid propagation may be incorrect. Instead, the plasmoid seems to have accelerated moving downtail. In the case-study of day 73 of 1993, a tailward-propagating plasmoid could not be detected on the basis of a bipolar signature of north-then-south deviation along the B z component of the magnetic ®eld because the spacecraft crosses the neutral sheet. Strong evidence of a plasmoid formation may be the isotropic angular distributions of energetic electrons at the CPS region detected during the ®rst transition from the north to the south lobe, as well as during the brief excursion to the CPS. The objection may arise that if the plasmoid formation has indeed taken place, then its passage over the spacecraft remaining at lobe magnetic ®eld lines should be manifested as the bipolar signature associated with a traveling compression region (TCR) . But we fail to determine such a distinct observational feature during the interval that spans the time between the reversal of energetic ion¯uxes.
The tailward propagation of a plasmoid has been considered the cause of the spacecraft motion from one lobe into the plasma sheet and then back into the same lobe again, as if a bulge in the plasma sheet had passed by . But here we did not have this typical case. Instead, it was the¯apping of the plasma sheet that resulted in entries of the spacecraft into the plasma sheet.
The energetic proton bursts at the PSBL are highly collimated along the magnetic ®eld and this is illustrated via the top angular distribution of Fig. 13 corresponding to the entry interval 16:56:02±16:57:45 UT into the plasma sheet of day 73, 1993. The ratio of tailward-to-Earthward¯ux is higher than three orders of magnitude and the vector magnetic ®eld deviates 7 o from Sunward. It is not observable any dawn-dusk anisotropy due to B´Ñj¯ow, where Ñj stands for the energetic-proton density gradient. Sarafopoulos and Sarris (1991a) used extensively the angular distributions as a powerful tool for probing the near-Earth PSBL stucture, but in the present case no one angular distribution upon the entry into the plasma sheet shows a detectable duskward¯ux (even if the distributions are scrutinized with time resolution as high as 3 s). Consequently, it is impossible to be estimated any spacecraft velocity relative to the PSBL. During the exit at $17:22 UT a gradient anisotropy of the just-receding layer is seen, but not so clearly as to derive a boundary velocity. The second average angular distribution of 58±77-keV protons shown in Fig. 13 corresponds to the CPS region from 17:05 to 17:15 UT of day 73, 1993. Although the particles continue streaming tailward, a less anisotropic character (as compared with that at the PSBL) is apparent. In this case the vector magnetic ®eld is highly variable throughout the interval and is thus eliminated.
The six successive PSBL crossings which occurred during the interval 16:45±18:20 UT of day 73, 1993, seem to be associated with a magnetospheric substorm having an onset just before the ®rst crossing. The combined eect of neutral-sheet¯apping and a tailwardmoving X-type magnetic ®eld neutral line is given via The Earthward¯uxes of energetic electrons and ions are associated with relatively low intensities of energetic electrons, which concurs with Richardson et al. (1993), who concluded that bursts of energetic electrons fall o with downtail distance such that they are nearly absent beyond $100 R E from the Earth. Also, contrary to expectations, in this work the Earthward ion¯uxes occurred with no positive B z component of the magnetic ®eld.
The inverse velocity dispersion (IVD) eect (i.e., lowenergy, $300-keV ions are detected 10±20 s before the high-energy, $1-MeV ions) observed with impulsive bursts of energetic particles inside the Earth's plasma sheet and PSBL (see Sarris and Axford, 1979;Sarafopoulos and Sarris, 1988;Taktakishvili et al., 1993), and interpreted as the growth rate of the accelerating source, must be disconnected with the velocity dispersions displayed in this work because of their striking dierence in time-scale. In addition, the IVDs occurred irrespectively of entries into and exits from the plasma sheet. Sarris et al. (1996) on one case-study of the velocity dispersion eect concluded that the sequence of events during the dispersive onset of energetic particle burst presented a self-consistent picture, being the result of time of¯ights from a semipermanent source (of energetic particles). In this study the incorporation of data sets with energies as low as $9.4-keV protons, in parallel with the fact that reverse velocity dispersions accompany exits from the plasma sheet, leads us to suggest the spatial structure of the plasma sheet as the cause producing the velocity dispersion eect.
In summary, the data presented here allow us to derive the following major ®ndings: 1. The energetic electrons are freely streaming almost exclusively at the PSBL and show peak¯uxes at lobe ®eld lines or at the PSBL. Conversely, when higher uxes of energetic electrons occur at CPS regions, then they show isotropic angular distributions, which are presumably associated with closed magnetic ®eld lines.
2. The ions are spatially separated in energy-dependent zones of tailward-jetting-particles. The lower energies occupy the innermost depths of the plasma sheet. A very accurate velocity ®lter eect has taken place.
3. The proton layer is con®ned within the energetic electron layer, the alpha particle layer is con®ned within the proton layer and the oxygen layer is con®ned within the alpha particle layer, if the same energy level for all the species is considered.
4. During the transition of the spacecraft from the north to the south lobe,``normal velocity dispersion'' is detected during the entry and``reverse velocity dispersion'' during the exit from the plasma sheet. The entire plasma sheet appears to display a``mirror-image'' in spatial structure, with the axis of symmetry being the neutral sheet.  The combined eect of neutral-sheet¯apping during the interval 16:45±18:20 UT of day 73, 1993 and a tailward-moving Xtype magnetic ®eld line is given in this sketch illustrating the GEOTAIL (S/C) trajectory. The six successive crossings of the PSBL occurred at $180 R E and are characterized by streaming ions in a direction emphasized by large arrows. The ®rst crossing occurred just after the substorm onset at $16:48 UT 5. We have determined the time-delay between the substorm onset on Earth and the ®rst detected energetic electron¯ux at $180 R E to be $4 min. The tailward-to-Earthward¯ow reversal occurred 82 min after the substorm onset. If we assume a constant plasmoid velocity then it is 205 km s A1 .